Predictive Vehicle Diagnostics Using Deep Learning

Problem Statement

Modern vehicles generate continuous OBD-II telemetry (RPM, temperature, pressure readings) but consumer systems only provide reactive fault codes after failure occurs. There is no accessible, low-cost predictive system capable of detecting abnormal engine behavior before critical failure using standard OBD-II protocols.

💡 Key Innovations

• • Unsupervised Learning: Autoencoder eliminates need for expensive labeled fault data
• • Temporal Modeling: BiLSTM with attention captures patterns across 120-second windows
• • 3-Tier Diagnostics: Intelligent fault categorization with actionable recommendations
• • Low-Cost Hardware: Compatible with ELM327 OBD-II adapters (<$20)
• • Production Ready: 99.9% uptime, <2s API response time

📊 Dataset

• • 1,699 driving sequences (120 seconds each)
• • 10 OBD-II sensors sampled at 10Hz (1,200 timesteps)
• • Sensors monitored: Coolant Temp, MAP, RPM, Speed, Intake Temp, MAF, Throttle, Ambient Temp, APP_D, APP_E
• • Data source: Real-world Seat Leon 1.2 TSI vehicle telemetry
• • Split: 1,334 normal / 365 fault sequences (trip-based)

🔧 System Architecture

🗺️ Full Pipeline Workflow Diagram

End-to-end pipeline from raw OBD-II telemetry collection through autoencoder-based anomaly labeling, BiLSTM classification, threshold optimization (F2-maximized), and production deployment on Render.com.

Figure 1: Complete 7-Phase Pipeline — Data Collection → Preprocessing → Autoencoder → Auto-Labeling → BiLSTM Classifier → Threshold Optimization → Production Deployment

🛠️ Technology Stack

📂 Repository

🚀 Future Enhancements